US restricts global access to Mythos class AI models after imposing export ban

Anthropic has suspended its latest Mythos 5 and Fable 5 artificial intelligence (AI) models, following a directive from US Commerce Secretary Howard Lutnick. The order, issued on Friday afternoon, instructed the company to restrict access for all foreign nationals due to national security concerns. A US official confirmed the order, which forced Anthropic to shut down access globally, including within the US.

The US government issued the directive after discovering a jailbreak that bypasses the Frontier model's built-in safety guardrails. Although the Trump administration had inspected the model for weeks and expressed concern about the potential exploitation of financial systems and critical infrastructure, a recent executive order stopped short of granting the power to block commercial releases.

Anthropic, which is currently in a legal battle with the US government, condemned the decision as a misunderstanding. The company argued in a blog article that a narrow jailbreak should not trigger a total commercial recall for hundreds of millions of users, warning that such a standard would halt industry-wide AI deployment. Furthermore, the company maintains that while the "government should have the ability to block unsafe deployments, as part of a statutory process that is transparent, fair, clear, and grounded in technical facts. This action does not adhere to those principles."

In response, a group of executives and technical leaders from the US and its allies have published an open letter to Secretary Lutnick and National Cyber Director Cairncross, calling for the export controls to be lifted and for the government to commit to an "open, scientific and transparent process of handling AI risk assessments in the future."

Meanwhile, an article in the Financial Times (£) argues that the directive cutting off global access to Mythos is a gift to China, saying that: "Washington has just done more to boost the appeal of Chinese AI models than Beijing could have ever hoped."

£ - The Financial Times article requires a subscription.

A version of this article is available without subscription in The Guardian.

Training Announcement: Freevacy offers a range of independently recognised professional AI governance qualifications and AI Literacy short courses that enable specialist teams to implement robust oversight, benchmark AI governance maturity, and establish a responsible-by-design approach across the entire AI lifecycle. Find out more.